Understanding DeepSeek's Influence
The debut of DeepSeek has generated significant waves in the technology sector, impacting hardware valuations and the race in artificial intelligence. This Chinese company specializes in creating open-source large language models (LLMs), and their recent announcements suggest a paradigm shift that has stirred the market considerably.
Let’s break down the essential details surrounding DeepSeek's introduction:
- DeepSeek has launched a new AI model called R1. This model parallels the popular O1 from OpenAI and is perceived as a formidable competitor to pricier American models.
- The development cost of the R1 model is touted to be around $6 million, a stark contrast to the hundreds of billions that US counterparts reportedly spend. However, some critics question the validity of this claim, noting reports that suggest DeepSeek had access to leading hardware like H100 chips. Nonetheless, this juxtaposition underscores a critical point: innovations in technology depend heavily on the preceding investments from US companies.
- The hardware sector, particularly companies like Nvidia (NASDAQ: NVDA), faced a harsh market reaction, with many stocks tumbling 20-30%. Investors feared that advancements like DeepSeek's R1 would diminish the need for high-end hardware. Yet, the reality is that even as companies leverage new models, they will increasingly require robust hardware to sustain advancements.
The AI Value Chain Disruption
The introduction of R1 is poised to disrupt the entire AI value chain. One critical insight that investors overlooked is that lower-priced models can open up a vast array of applications for AI technologies.
Often likened to Jevon’s Paradox—where increased efficiency leads to greater consumption—this principle applies well to AI's evolving landscape. Essentially, as products become more affordable, usage becomes more widespread, enhancing opportunities across various sectors.
Revolutionizing AI Applications
For AI applications, dematerializing costs can lead to tremendous growth potential.
Both O1 and R1 models are classified as "test-time scaling" models. This innovation implies that they engage in preliminary processing before delivering outputs. Such advancements are vital for an emerging facet of AI known as Agentic AI.
Agentic AI is akin to currently used copilots; however, it possesses the ability to engage in deeper reasoning compared to conventional AI chatbots that simply generate responses based on user prompts. For more context, we discussed Agentic AI in earlier reports.
The reach of Agentic AI encompasses numerous service sectors, including customer service, content generation, and healthcare. Thus, companies with extensive software user bases stand to gain significantly from these developments.
- The software development arena was among the first to adopt copilot methodologies, with firms like GitLab (NASDAQ: GTLB) and GitHub leading the charge.
- Similarly, productivity platforms are not left behind, with major players like Microsoft (NASDAQ: MSFT) and Salesforce (NYSE: CRM) enhancing their offerings through co-pilot products.
- Cybersecurity firms such as CrowdStrike (NASDAQ: CRWD) and Datadog (NASDAQ: DDOG) also increasingly integrate AI capabilities to boost their service offerings.
Market projections suggest that the potential valuation of Agentic AI could exceed one trillion dollars within the next decade, a topic we elaborated on in our latest webinar.
Open-Source vs. Closed-Source Models
The release of DeepSeek's model presents challenges particularly for companies that offer closed-source models. The open-source approach is gaining traction, creating formidable competition for proprietary models.
With innovations stemming from the DeepSeek model now accessible to all model providers, we can expect many companies to incorporate these advancements into their future offerings.
Open-source companies have alternative monetization strategies that provide them with an edge (e.g., Meta (NASDAQ: META)). DeepSeek’s ownership is also involved in other ventures including a quant hedge fund, hinting that the financial benefits of such innovations could extend far beyond the initial development costs.
The Future of Compute Hardware and Energy Solutions
While the AI applications sector stands to gain from these developments, the hardware landscape has witnessed challenges, leading to significant investment opportunities due to recent stock declines.
There are three primary opportunity sectors that investors should consider:
- Compute — Currently dominated by Nvidia, with predictions of growth in custom chip manufacturing by the year 2025.
- Networking — A sector poised for innovation and new players seizing opportunities.
- Power Generation — This industry, which has been underfunded for over 20 years, is anticipated to experience a revival.
Despite recent struggles, major technology firms are increasing their hardware investment, reflecting confidence in future growth. For instance, Meta has announced a significant year-over-year increase in capital expenditure.
Frequently Asked Questions
What is DeepSeek?
DeepSeek is a Chinese artificial intelligence company known for developing open-source large language models.
What is the R1 model developed by DeepSeek?
R1 is a new AI model from DeepSeek that competes with models like OpenAI's O1.
How does the release of R1 impact AI applications?
The introduction of R1 is likely to lower operational costs, enhancing accessibility and usage of AI applications across industries.
What is Agentic AI?
Agentic AI refers to advanced AI that can perform reasoning tasks, compared to traditional chatbots that operate primarily through user prompts.
How are hardware companies affected by DeepSeek’s innovations?
Hardware companies, especially Nvidia, faced declines in their stock prices due to fears of reduced demand for high-end hardware.
